Agent skill

tune-detection

Tune exercise detection sensitivity and thresholds. Use when exercises aren't counting correctly, have false positives, or need calibration adjustments.

Stars 163
Forks 31

Install this agent skill to your Project

npx add-skill https://github.com/majiayu000/claude-skill-registry/tree/main/skills/data/tune-detection

SKILL.md

Tune Detection Thresholds

Common Issues

Problem Likely Cause Fix
Not counting reps Thresholds too strict Lower down threshold or raise up threshold
Double counting Thresholds too loose Tighten thresholds, add hysteresis
Counts on wrong motion Wrong landmarks Check landmark IDs match exercise
Works for some people Fixed thresholds Use body-relative thresholds with getBodyScale()

Threshold Locations

JSON configs (preferred):

exercises/*.json → detection.thresholds

Legacy functions in exercise_ui.html:

  • detectLegacySquat - DOWN_ANGLE, UP_ANGLE
  • detectLegacyPushup - DOWN_ANGLE, UP_ANGLE
  • detectLegacyJumpingJack - arm position relative to shoulders
  • detectLegacyCalfRaise - legLength * 0.04 threshold
  • detectLegacySideStretch - bodyScale * 0.4 threshold

Detection Helpers

The codebase includes helpers for robust detection:

javascript
// Check if landmarks are visible
isVisible(landmarks, [23, 24, 25, 26], 0.5)

// Normalize to body size
const bodyScale = getBodyScale(landmarks);
const threshold = baseThreshold * bodyScale;

Testing Changes

  1. Edit threshold in JSON or HTML
  2. Restart tracker: ./exercise_tracker.py user_prompt_submit '{}'
  3. Watch the status text - it shows live angle/distance values
  4. Adjust based on when state transitions happen

Both-sides Averaging

For angle-based exercises, use joint_alt in JSON to average both sides:

json
"landmarks": {
  "joint": [23, 25, 27],
  "joint_alt": [24, 26, 28]
}

Expand your agent's capabilities with these related and highly-rated skills.

Didn't find tool you were looking for?

Be as detailed as possible for better results